摘要
基于深度学习的信道估计方法中,训练网络模型需要大量的数据运算,且所有用户数据都需要集中上传至服务器上,存在隐私泄漏的隐患。针对上述问题,提出了一种基于联邦学习的LTE-V2X(Long Term Evolution-Vehicle to Everything)信道估计算法,采用CNN-LSTM-DNN(Convolutional Neural Network-Long Short Term Memory-Deep Neural Network)模型对时变的信道进行估计,并将学习网络模型所需要的计算分配到车载用户中,在降低道旁基站负载的同时也保护了车载用户数据的隐私。仿真结果表明,基于联邦学习的信道估计算法在车载用户高速移动的场景下,较传统的信道估计算法平均有10 dB以上的归一化均方误差(Normalized Mean Square Error,NMSE)增益以及3 dB以上的误码率(Bit Error Rate,BER)增益,且较集中式学习算法相比,NMSE性能差距在3 dB以内;BER性能差距在1 dB以内,所提算法能够有效追踪时变的信道,且与集中式学习算法相比仅损失了极少的性能。
In channel estimation methods based on deep learning,training network models requires a lot of data operations,and all user data needs to be uploaded to the server in a centralized manner,which poses a hidden danger of privacy leakage.In response to above problems,a long term evolution-vehicle to everything(LTE-V2X)channel estimation algorithm based on federated learning is proposed,which uses the convolutional neural network-long-short term memory-deep neural network(CNN-LSTM-DNN)model to predict the time-varying channel,and allocates the calculations required to learn the network model to the vehicle users.While reducing the load of roadside base stations,it also protects the privacy of vehicle users'data.The simulation results show that the channel estimation algorithm based on federated learning has an average normalized mean square error(NMSE)gain of more than 10 dB and a bit error rate(BER)gain of more than 3 dB compared with the traditional channel estimation algorithm in the scene of high-speed movement of vehicle users.Compared with the centralized learning algorithm,The NMSE performance gap is within 3 dB,and the BER performance gap is within 1 dB.The proposed algorithm can effectively track time-varying channels,and the centralized learning algorithm only loses very little performance.
作者
景兴红
尹子松
蔡志镕
何世彪
廖勇
JING Xinghong;YIN Zisong;CAI Zhirong;HE Shibiao;LIAO Yong(School of Electronic Information,Chongqing Institute of Engineering,Chongqing 400056,China;School of Microelectronics and Communication Engineering,Chongqing University,Chongqing 400044,China)
出处
《电讯技术》
北大核心
2021年第6期681-688,共8页
Telecommunication Engineering
基金
国家自然科学基金资助项目(61501066)
重庆市自然科学基金资助项目(cstc2019jcyj-msxmX0017)。
关键词
车联网
信道估计
深度学习
联邦学习
Internet of Vehicles
channel estimation
deep learning
federated learning
